The accuracy of burn diagnosis codes in health administrative data: A validation study. Issue 2 (March 2017)

• Record Type: Journal Article
• Authors: Mason, Stephanie A.; Nathens, Avery B.; Byrne, James P.; Fowler, Rob; Gonzalez, Alejandro; Karanicolas, Paul J.; Moineddin, Rahim; Jeschke, Marc G.
• Is Part Of: Burns. Volume 43:Issue 2(2017)
• Page Start: 258
• Page End: 264
• Publication Date: 2017-03
• Subjects: Burns -- Burn -- Validation -- Administrative data; Burns and scalds -- Periodicals; 617.11
• DOI: 10.1016/j.burns.2016.11.008
• Languages: English
• ISSNs: 0305-4179

Abstract

Highlights: A clinical burn registry was linked to an administrative database to assess the accuracy of ICD-10 diagnosis codes. Codes for burn extent are highly sensitive and specific in identifying patients with ≥10% and ≥20% TBSA injury. Codes for inhalation injury and % full-thickness TBSA are less reliable with lower sensitivity and specificity. Abstract: Background: Health administrative databases may provide rich sources of data for the study of outcomes following burn. We aimed to determine the accuracy of International Classification of Diseases diagnoses codes for burn in a population-based administrative database. Methods: Data from a regional burn center's clinical registry of patients admitted between 2006–2013 were linked to administrative databases. Burn total body surface area (TBSA), depth, mechanism, and inhalation injury were compared between the registry and administrative records. The sensitivity, specificity, and positive and negative predictive values were determined, and coding agreement was assessed with the kappa statistic. Results: 1215 burn center patients were linked to administrative records. TBSA codes were highly sensitive and specific for ≥10 and ≥20% TBSA (89/93% sensitive and 95/97% specific), with excellent agreement ( κ, 0.85/ κ, 0.88). Codes were weakly sensitive (68%) in identifying ≥10% TBSA full-thickness burn, though highly specific (86%) with moderate agreement ( κ, 0.46). Codes for inhalation injury had limited sensitivity (43%) but high specificity (99%) with moderate agreement ( κ, 0.54). Burn mechanism had excellent coding agreement ( κ, 0.84). Conclusions: Administrative data diagnosis codes accurately identify burn by burn size and mechanism, while identification of inhalation injury or full-thickness burns is less sensitive but highly specific.